Correction for bias in visual transect censuses of coral reef fishes

Transect techniques for censusing reef fishes, and the sources of bias inherent in them are considered. A technique, derived from aeraly survey methods, is demonstrated to correct a bias in density estimates due to the width of the transect being censused. This bias is sufficient on a transect 1 m wide to underestimate density by 11.1–26.7% for five species or species groups examined. The bias is still greater on wider transects. Because this bias varies in degree among species, comparisons among species should not be made using uncorrected transect data. Comments are made on other probable sources of bias in transect data, and on ways of minimising bias when making visual transect censuses.     

An Evaluation of Survey Methods for Estimating Northern Bobwhite Abundance in Southern Texas

Abstract: Distance sampling has been identified as a reliable and well-suited method for estimating northern bobwhite (Colinus virginianus) density. However, distance sampling using walked transects requires intense sampling to obtain precise estimates, thus making the technique impractical for large acreages. Researchers have addressed this limitation by either resorting to the use of indices (e.g., morning covey-call surveys) or incorporating the use of aerial surveys with distance sampling. Both approaches remain relatively untested. Our objectives were to 1) compare density estimates among morning covey-call surveys, helicopter transects, and walked transects; 2) test a critical assumption of distance sampling pertinent to helicopter surveys (i.e., all objects on line are detected); and 3) evaluate the underlying premise of morning covey-call surveys (i.e., that the no. of calling coveys correlates with bobwhite density). Our study was conducted on 3 study sites in Brooks County, Texas, USA, during October to December, 2001 to 2005. Comparisons between walked transects and morning covey-call surveys involved the entire 5-year data set, whereas helicopter transects involved only the latter 2 years. Density estimates obtained from helicopter transects were similar to walked transect estimates for both years. We documented a detection probability on the helicopter transect line of 70 ± 10.2% (% ± SE; n = 20 coveys). Morning covey-call surveys yielded similar density estimates to walked transect estimates during only 2 of 5 years, when walked transect estimates were the least accurate and precise. We detected a positive relationship (R² = 0.51; 95% CI for slope: 29.5–53.1; n = 63 observations) between covey density and number of coveys heard calling. We conclude that helicopter transects appear to be a viable alternative to walked transects for estimating density of bobwhites. Morning covey-call surveys appear to be a poor method to estimate absolute abundance and to depict general population trajectories.     

Comparative performance of four survey methods for assessing Lilian’s Lovebird abundance in Liwonde National Park,Malawi

Monitoring abundance of threatened species is important for conservation planning. Lilian’s Lovebird Agapornis lilianae is a near-threatened small parrot found in mopane Colophospermum mopane woodland. Its population has not been investigated in any part of its range. We investigated the abundance and density of the Lilian’s Lovebird in Liwonde National Park, Malawi. Both distance sampling (line and point transects) methods and total counts (waterhole and flyway counts) were applied. The point count method gave very low numbers and was discontinued after the first year. Line transects conducted during the wet season had the highest density estimates of 17 ± 4.8 lovebirds km?2 of mopane woodland. However, number of observations per transect in each year were low. Waterhole counts had the lowest density estimates (10 ± 3.5 lovebirds km?2). Flyway counts gave an intermediate estimate (13 ± 3.0 lovebirds km?2). The total population of Lilian’s Lovebirds in Liwonde National Park is therefore estimated to be about 4 000 individuals. The use of line transect counts at the end of the rainy season is recommended for continued monitoring of Lilian’s Lovebirds abundance in Liwonde National Park.     

The Influence of Transect Use by Local People and Reuse of Transects for Repeated Surveys on Nesting in Western Lowland Gorillas (Gorilla gorilla gorilla) and Central Chimpanzees (Pan troglodytes troglodytes) in Southeast Cameroon

Monitoring populations of endangered species over time is necessary to guide and evaluate conservation efforts. This is particularly important for nonprotected areas that ensure connectivity between protected populations but are prone to uncontrolled hunting pressure. We investigated whether use of transects by local people and transect reuse for repeated surveys influence great ape nesting and bias results. We conducted simultaneous marked nest count surveys over 12 mo on established and newly opened transects in a nonprotected area subject to traditional heavy use by local people and recorded forest composition and signs of human activity. Chimpanzee and gorilla density estimates and encounter rates per kilometer were lower on established transects than on new ones. A generalized linear model indicated that hunting activity, distance to a regularly used forest trail, and transect type (old or new) predicted chimpanzee nest abundance, and distance to the trail and transect type predicted gorilla nest abundance, with no effect of habitat type (percentage suitable habitat) for either species. We, therefore, suggest that the difference in great ape nesting is a result of high levels of hunting by local people on established transects and forest trails. Our results support the use of repeated line transect surveys for monitoring great ape populations in many circumstances, although we advocate taking precautions in nonprotected areas, to avoid the bias imposed by use of established transects for hunting.     

A simple alternative to line transects of nests for estimating orangutan densities

We conducted a validation of the line transect technique to estimate densities of orangutan (Pongo pygmaeus) nests in a Bornean swamp forest, and compared these results with density estimates based on nest counts in plots and on female home ranges. First, we examined the accuracy of the line transect method. We found that the densities based on a pass in both directions of two experienced pairs of observers was 27% below a combined sample based on transect walks by eight pairs of observers, suggesting that regular line-transect densities may seriously underestimate true densities. Second, we compared these results with those obtained by nest counts in 0.2-ha plots. This method produced an estimated 15.24 nests/ha, as compared to 10.0 and 10.9, respectively, by two experienced pairs of observers who walked a line transect in both directions. Third, we estimated orangutan densities based on female home range size and overlap and the proportion of females in the population, which produced a density of 4.25–4.5 individuals/km² . Converting nest densities into orangutan densities, using locally estimated parameters for nest production rate and proportion of nest builders in the population, we found that density estimates based on the line transect results of the most experienced pairs on a double pass were 2.82 and 3.08 orangutans/km², based on the combined line transect data are 4.04, and based on plot counts are 4.30. In this swamp forest, plot counts therefore give more accurate estimates than do line transects. We recommend that this new method be evaluated in other forest types as well.     

Determining Methods of Underwater Visual Census for Estimating the Abundance of Coral Reef Fishes

We aimed to determine optimal methods of underwater visual census (UVC) for estimating the abundance of coral reef fishes exploited by fisheries in the tropical Pacific. Two main methods were tested using SCUBA: strip transect and stationary point counts. We assessed their relative accuracy, precision, power and efficiency (cost), and compared different census area dimensions, observer swimming speeds, and number of replicates. Twenty-five comparisons of the two UVC methods were conducted for 73 species from five families of coral reef fishes, on reefs in Australia and Fiji. Species were grouped within families based on their mobility. Few significant differences were found, either among or between strip transects and stationary point counts. The data were characterised by high variability, low precision and low power. A trend for greater accuracy in density estimates of small sedentary species with smaller census area was apparent, which probably reflects searching efficiency. Only one species group, the sedentary Acanthuridae showed differences between transects and point counts. Higher, hence presumably more accurate, density estimates were obtained with 50m×5m transects. Notably, point counts could be deployed in 70% of the time of transects. A bootstrapping procedure demonstrated a consistent improvement in precision of density estimates with increasing number of replicates, but no appreciable change in precision was found beyond 10 to 15 replicates, in all species groups and for both transects and point counts. Consequently, and because of the high variability inherent in fish density estimates, we recommend that at least 10 replicates be used to quantify the species considered here. The power calculations showed that only large differences in density will be detected with the replication levels typical of UVC surveys. Power was greatest for the roving serranids, with a decrease in density of about 50% detectable for this species group. We discuss the importance of carefully defining the behavioural attributes of species prior to selecting a UVC method. A method of temporal stratification in a count is described for censusing a range of species of varying mobilities. We suggest that fish mobility and search efficiency are key factors in optimising UVC methods.     

The use of block counts,mark-resight and distance sampling to estimate population size of a mountain-dwelling ungulate

Population size estimates represent indispensable tools for many research programs and for conservation or management issues. Mountain ungulates in open areas are often surveyed through ground counts that normally underestimate population size. While the use of sample counts is desirable, few studies have compared different probabilistic approaches to estimate population size in this taxon. We compared the size estimates of a male population of Alpine chamois using mark-resight and line transect sampling methods, while block counts were used to obtain the minimum number of males alive in the study area. Surveys were conducted within the Gran Paradiso National Park (Italy), in August–September 2013, using block counts along purposely selected trails and vantage points, mark-resight over 5 consecutive resightings from vantage points and trails, and line transect sampling along 12 transects repeated 8 times. Block counts yielded a minimum number of males alive in the population of n = 72 individuals. This value was greater than the upper bound of the 95 % confidence interval achieved using line transect sampling {n = 54, CV = 14 % [95 % CI (40, 71)]} while mark-resight yielded a more realistic result of n = 93 individuals {CV = 18 % [95 % CI (63, 137)]}. Our results suggest that line transect sampling performed poorly in the Alpine environment, leading to underestimates of population size, likely due to violations of some assumptions imposed by the rugged nature of the terrain. The mark-resight yielded lower precision, possibly due to the limited number of marked individuals and resighting occasions, but it provided robustness and accurate estimates as marks were evenly distributed among animals.     

Biases associated with the use of underwater visual census techniques to quantify the density and size-structure of fish populations

Estimates of reef fish densities made by divers visually censusing 5 m wide strip transects were compared with capture-resight estimates calculated independently using data on the resighting frequencies of fish marked with colour-coded tags. The difference in density estimates between methods varied between species but with patterns consistent at the three eastern Tasmanian sites studied. Densities of the two most abundant species, the wrasses Notolabrus tetricus and Notolabrus fucicola, showed good agreement between methods. By contrast, populations of the two monacanthids Meuschenia australis and Meuschenia freycineti were underestimated by an order of magnitude in strip transects relative to capture-resight, while populations of the open-water latrid Latridopsis forsteri were overestimated.For all common fish species observed in strip transects, variation in density estimates between divers was extremely low compared to variation between sites and between months. Variation in density estimates between different days within a week and between 50 m blocks located 50 m apart was also negligible; however, residual error associated with variation by a single diver within a block and day was high. Biases associated with changing detectability of fishes in open versus densely vegetated habitats were assessed by algal removal experiments, and were found to be negligible or non-significantly low for five of the six species examined. The exceptional species—N. tetricus—exhibited a 70% rise in fish sighted by diver following algal clearance despite trapping data indicating no increase in fish numbers.Diver estimates of fish length were on average 7% greater than measured lengths. Divers possessed a clear tendency to make increasingly inaccurate size estimates as fish length deviated in either direction from 300 mm. Lengths of 175-mm animals were underestimated by ≈20% and 400-mm fishes were overestimated by ≈10%. These changes in diver bias with fish size were largely independent of fish species, site and diver. The precision of size estimates, as indicated by the standard deviation of bias, also varied with fish size, with values varying from ≈13% at 200 mm fish length to ≈8% at 400 mm length. The decline in precision at small body size largely reflected size intervals used by divers to bin data. Divers appear capable of making more precise size estimates than the 25-mm interval used at small fish sizes.     

Features of Species and Structural Diversity of Artificial Pine Stands on Soddy–Podzolic Loam Soils

The method of ecological transects was used for studying plant-species composition in an aged artificial pine stand created on a composite spruce forest on soddy–podzolic soil (underlain by blanket loams) in the broadleaved–spruce forest subzone (Moscow oblast). The change of edificatory tree species was followed by changes in the ecological conditions and, as a consequence, in the species composition, abundance, and cenotic significance of vegetation. Changes in the pattern of restoration of floristic diversity in different soil-geochemical facies of the transect were observed.     

Line Transect and Triangulation Surveys Provide Reliable Estimates of the Density of Kloss’ Gibbons (Hylobates klossii) on Siberut Island, Indonesia

Estimating population densities of key species is crucial for many conservation programs. Density estimates provide baseline data and enable monitoring of population size. Several different survey methods are available, and the choice of method depends on the species and study aims. Few studies have compared the accuracy and efficiency of different survey methods for large mammals, particularly for primates. Here we compare estimates of density and abundance of Kloss’ gibbons (Hylobates klossii) using two of the most common survey methods: line transect distance sampling and triangulation. Line transect surveys (survey effort: 155.5 km) produced a total of 101 auditory and visual encounters and a density estimate of 5.5 gibbon clusters (groups or subgroups of primate social units)/km². Triangulation conducted from 12 listening posts during the same period revealed a similar density estimate of 5.0 clusters/km². Coefficients of variation of cluster density estimates were slightly higher from triangulation (0.24) than from line transects (0.17), resulting in a lack of precision in detecting changes in cluster densities of <66 % for triangulation and <47 % for line transect surveys at the 5 % significance level with a statistical power of 50 %. This case study shows that both methods may provide estimates with similar accuracy but that line transects can result in more precise estimates and allow assessment of other primate species. For a rapid assessment of gibbon density under time and financial constraints, the triangulation method also may be appropriate.     

Improving Aquatic Warbler Population Assessments by Accounting for Imperfect Detection

Monitoring programs designed to assess changes in population size over time need to account for imperfect detection and provide estimates of precision around annual abundance estimates. Especially for species dependent on conservation management, robust monitoring is essential to evaluate the effectiveness of management. Many bird species of temperate grasslands depend on specific conservation management to maintain suitable breeding habitat. One such species is the Aquatic Warbler (Acrocephalus paludicola), which breeds in open fen mires in Central Europe. Aquatic Warbler populations have so far been assessed using a complete survey that aims to enumerate all singing males over a large area. Because this approach provides no estimate of precision and does not account for observation error, detecting moderate population changes is challenging. From 2011 to 2013 we trialled a new line transect sampling monitoring design in the Biebrza valley, Poland, to estimate abundance of singing male Aquatic Warblers. We surveyed Aquatic Warblers repeatedly along 50 randomly placed 1-km transects, and used binomial mixture models to estimate abundances per transect. The repeated line transect sampling required 150 observer days, and thus less effort than the traditional ‘full count’ approach (175 observer days). Aquatic Warbler abundance was highest at intermediate water levels, and detection probability varied between years and was influenced by vegetation height. A power analysis indicated that our line transect sampling design had a power of 68% to detect a 20% population change over 10 years, whereas raw count data had a 9% power to detect the same trend. Thus, by accounting for imperfect detection we increased the power to detect population changes. We recommend to adopt the repeated line transect sampling approach for monitoring Aquatic Warblers in Poland and in other important breeding areas to monitor changes in population size and the effects of habitat management.     

Comparison of survey methods for wintering grassland birds

ABSTRACT Although investigators have evaluated the efficacy of survey methods for assessing densities of breeding birds, few comparisons have been made of survey methods for wintering birds, especially in grasslands. In winter, social behavior and spatial distributions often differ from those in the breeding season. We evaluated the degree of correspondence between density estimates based on different survey methods. Surveys were conducted during two winters (2001–2002 and 2002–2003) on 16 grassland sites in southwestern Oklahoma. Line‐transect (using a detection function to account for birds present but not detected) and area‐search (where density was based on the total count within a given area) methods were employed. Observations on line transects were also analyzed as strip transects, where density was based on total count within a given strip width and no detection function was used. Savannah Sparrows (Passerculus sandwichensis), LeConte's Sparrows (Ammodramus leconteii), Song Sparrows (Melospiza melodia), Smith's Longspurs (Calcarius pictus), Chestnut‐collared Longspurs (C. ornatus), and Eastern Meadowlarks (Sturnella magna) were sufficiently abundant to allow comparison. Area‐search density estimates tended to be higher than line‐transect estimates for Savannah Sparrows, Song Sparrows, and Eastern Meadowlarks, suggesting that some individuals initially located close to the transect line were not detected on line transects. The area‐search and line‐transect methods gave similar density estimates for Chestnut‐collared and Smith's longspurs. Area‐search estimates of Eastern Meadowlarks were significantly higher in the second year of the study only. For this species, area‐search estimates did not differ from those of strip transects covering an equal area, so the reason for the differing meadowlark estimates is not clear. Higher density estimates using the area‐search method likely resulted from: (1) birds that might escape detection by hiding were more likely detected (flushed) during area searches because of the repeated passes through the area, and (2) birds close to the line in line transects escape detection by hiding, biasing those estimates low. We also evaluated the correspondence of density rankings for the six species as determined by the different survey methods and for the same species across survey sites. Correlations among the six species of the area‐search results with those of line transects and strip transects generally were high, increasing in 2002–2003 when densities of birds were greater. All three methods provided similar density rankings among species. Density rankings within species across sites for the four non‐longspur species generally were concordant for the three methods, suggesting that any of them will adequately reflect among‐site differences, especially when densities vary greatly across sites. Further research is needed to determine the extent to which grassland birds are missed on line transects. We suggest that workers using line transects to study these species give careful consideration and make additional efforts to satisfy the distance‐sampling assumption that all birds on or near the line are detected. If density is measured as a total count in a fixed area, we recommend that observers pass within <10 m of all points in the area.     

Sources of Variability in the Measurement of Fungal Spore Yields

Variability in the production of fungal spores and in the measurement of spore yields was investigated in four species of fungi: Colletotrichum gloeosporioides, Colletotrichum coccodes, Colletotrichum phomoides, and Acremonium strictum. When the fungi were grown on solid medium in microplates and spore yields were measured by counting the subsamples with a hemacytometer, the variability among hemacytometer squares was always the largest source of variation, accounting for 51 to 91% of the total variation. Variability among replicate cultures and results of repeat experiments were generally also significant. The effect of square-to-square variability on the precision of spore yield measurement was minimized by counting a moderate number (ca. 30) of squares per culture. Culture-to-culture variability limited the practical precision of spore production measurements to a 95% confidence interval of approximately the mean ± 25%. We provide guidelines for determining the number of replicate cultures required to attain this or other degrees of precision. Particle counter-derived spore counts and counts based on spore weights were much less variable than were hemacytometer counts, but they did not improve spore production estimates very much because of culture-to-culture variability. Results obtained by both of these methods differed from those obtained with a hemacytometer; particle counter measurements required a correction for spore pairs, while the relationship between spore weights and spore counts changed as the cultures aged.     

The line transect method for estimating densities of large mammals in a tropical deciduous forest: An evaluation of models and field experiments

We have evaluated techniques of estimating animal density through direct counts using line transects during 1988–92 in the tropical deciduous forests of Mudumalui Sanctuary in southern India for four species of large herbivorous mammals, namely, chital (Axis axis). sambar (Cervus unicolor). Asian elephant (Elephas maximus) and gaur (Bos gaurus) Density estimates derived from the Fourier Series and the Half-Normal models consistently had the lowest coefficient of variation. These two models also generated similar mean density estimates. For the Fourier Series estimator, appropriate cut-off widths for analyzing line transect data for the four species are suggested. Grouping data into various distance classes did not produce any appreciable differences in estimates of mean density or their variances, although model fit is generally better when data arc placed in fewer groups. The sampling effort needed to achieve a desired precision (coefficient of variation) in the density estimate is derived. A sampling effort of 800 km of transects returned a 10% coefficient of variation on estimate for ehital; for the other species a higher effort was needed to achieve this level of precision. There was no statistically significant relationship between detectability of a group and the size of the group for any species. Density estimates along roads were generally significantly different from those in the interior of the forest, indicating that road-side counts many not be appropriate for most species.     

Estimating phytoplankton carbon from microscopic counts: an application for riverine systems

Algal biomass, in addition to cell numbers, is a measure of the successful conversion of inorganic to organic carbon. Consequently, carbon is the main currency used in aquatic models and in flux and budget studies. On the other hand, microscopic observation and counts remain the only means for determining species composition and biomass, which is relevant to many aspects of aquatic ecology. In this study, we focus on the way to convert biovolume to carbon biomass for algal assemblages of two rivers, using a computerized system that records dimensions of phytoplankton (Gosselain & Hamilton, 2000). We first compare different equations found in the literature for converting algal cell volume to cellular carbon content. We then evaluate the accuracy of a biomass estimate based on less time-consuming measurements, using pre-determined biovolume values instead of measuring cells in all samples. Biovolume/carbon equations are evaluated using total phytoplankton carbon biomass determined from measured chlorophyll a. Equations established for freshwater taxa seem to provide better estimates of algal biomass in the two case studies presented here, the Rideau and Meuse rivers (Canada and Belgium, respectively) than do more numerous equations defined for marine taxa. Furthermore, equations that make a distinction between diatoms and other algae appear more appropriate than those considering all algal groups as a whole. Finally, mean values of algal biovolumes, determined using sufficient measurements of cell dimensions from representative sampling series, may prove sufficient for carbon estimates of taxa in relatively homogenous size ranges. The careful choice of appropriate volumetric shapes and taxa categories remains of prime importance to get precise results.     

Quality control of Utermöhl-based phytoplankton counting and biovolume estimates—an easy task or a Gordian knot?

The most suitable way for standardizing biovolume based assemblage analysis using the Utermöhl technique for counting combined with biomass conversion has not yet been found even though this method has been successfully used in plankton studies for many years. Due to the complexity of the approach easily applicable steps to initially meet primary standard end point or target counts for intra-laboratory standardization tested here seem promising. Examples from count validation and biometric data from the large datasets of three laboratories are given. The first two examples initially intended to quantify the taxon specific scatter of counts by (A) using identical replica of field samples combined with a half-chamber count (scan of every second transect) and the 30 random field approach, respectively, and (B) the replication of transect counts in one sample. Both examples identified relatively low minimum count thresholds to delimit counting errors for single cell counts. The third example identifies shape variability and allometric relationships of the main axes and shows a way to improve taxon specific biomass estimates with special reference to cell thickness. However improved precision of quantitative phytoplankton analysis requires optimization of combined counting strategy and biovolume assessment methods.     

鸟类调查的样线法和样点法比较:以崇明东滩春季盐沼鸟类调查为例

为了探讨样线法和样点法对盐沼湿地鸟类调查的有效性及适用性, 我们于2008年3–5月采用固定距离样线法和固定半径样点法对崇明东滩盐沼鸟类进行了调查。5次调查共记录到鸟类24种, 其中样线法记录到19种, 样点法记录到21种, 两种调查方法平均每次记录到的鸟类种数没有显著差异, 表明两种方法对盐沼鸟类种类的发现能力相似。两种方法得到的鸟类密度之间有显著的相关性,说明两种方法在反映鸟类群落中每种鸟类的相对数量方面具有可比性。但除了仅在一种调查方法中记录到的种类, 样点法调查得到的每种鸟类密度和鸟类总密度均高于样线法的调查结果。因此, 在对鸟类种群或群落的时空特征进行比较时, 需要考虑所采用的调查方法对调查结果的影响。     

Comparative analysis of visual census techniques for highly mobile,reef-associated piscivores (Carangidae)

Synopsis Visual census techniques applicable to coral reef-associated fishes are reviewed and the results of field tests using six (three transect-based and three point-based) to estimate the density of carangids at Carter Reef, Great Barrier Reef, are presented. Data are analyzed with respect to the effects of observers on fishes seen, observer biasses, precision of the estimates and, as far as possible, accuracy of the estimates. Transects generate estimates of population density and structure different from those of point-based estimates. Various point-based census methods, however, generate density estimates consistent with one another and are generally more precise than transect-based methods. The results of the field study obviously cannot be generalized to other quite different types of reef fishes. The problems we encountered and a review of the techniques used to census reef fishes visually in the past, however, suggest that: (1) interval counts, such as Rapid Visual Census techniques, are likely to be inaccurate and difficult to compare; (2) for species with high probabilities of detection, instantaneous area counts appear to be the most effective way to estimate densities, whereas cryptic species are best censused using instantaneous variable distance point counts, and (3) strip transects may often be less efficient than line transects, due to inconstant levels of subject detectability.     

The differences of butterfly (Lepidoptera, Papilionoidea) communities in habitats with various degrees of disturbance and altitudes in tropical forests of Vietnam

A survey was conducted on the species composition, richness and abundance of Papilionoidea (excluding Lycaenidae) butterfly fauna in habitats with various degrees of disturbance and altitudes in tropical forests at Tam Dao National Park, northern Vietnam in 2001. The transect method was used to collect data in the survey. Six transects representing different habitat types at two sites, one site located at a low elevation of 200–250 m a.s.l., and the other located at a high elevation of 950–1000 m a.s.l., were chosen: three transects for each site, with a length of 500 m for each transect. A total of 3594 individuals of 127 species in 240 sets of data were recorded from various habitats. The differences in butterfly composition, species richness, abundance and diversity in different habitat types and altitudes were analyzed. The results showed significant differences of butterfly diversity among the different habitat types and between the low and high altitude sites. The butterfly diversity, species richness and species abundance in the low elevation habitats were higher than in the high elevation habitats. The highest diversity of butterflies occurred in the mixed habitats of agriculture, scrub and clearing lands of high disturbance. However, butterflies most important for conservation are associated with undisturbed or moderately disturbed forests only.     

Phytoplankton composition in the Weddell-Scotia Confluence area during austral spring in relation to hydrography

During the EPOS leg 2 cruise of the RV Polarstern, carried out in late austral spring of 1988–1989, the composition of phytoplankton in relation to the distribution of hydrographic parameters was studied in four successive transects carried out along 49°W and 47°W, across the Weddell-Scotia Confluence (WSC) and the marginal ice zone (which overlapped in part). In all transects, a maximum of phytoplankton biomass was found in the WSC, in surface waters stabilized by ice melting. Different phytoplankton assemblages could be distinguished. North of the Scotia Front (the northern limit of the WSC) diatoms with Chaetoceros neglectus, Nitzschia spp. and (Thalassiosira gravida) dominated the phytoplankton community. This assemblage appeared to have seeded a biomass maximum which occupied, during the first transect, an area of the WSC, south of the Scotia Front. The southernmost stations of the first transect and all the stations to the south of the Scotia Front in the other transects were populated by a flagellate assemblage (with a cryptomonad, Pyramimonas spp. and Phaeocystis sp.) and an assemblage of diatoms (Corethron criophilum and Tropidoneis vanheurkii among others) associated to the presence of ice. During the last three transects, the flagellate assemblage formed a bloom in the low salinity surface layers of the WSC zone. The bulk of the biomass maximum was formed by the cryptomonad which reached concentrations up to 4×10⁶ cells l^–1 towards the end of the cruise. Multivariate analysis is used to summarize phytoplankton composition variation. The relationships between the distribution of the different assemblages and the hydrographic conditions indicate that the change of dominance from diatoms to flagellates in the WSC zone was related to the presence of water masses from different origin.Data presented here were collected during the European Polarstern Study (EPOS) sponsored by the European Science Foundation